Designing for a Future We Cannot Predict: The New Science of Resilient Infrastructure

Engr. Kamran Abbas

BSc Civil Engineering

MS Transportation Engineering

Introduction
The Problem With Designing for a Fixed Future
Why Historical Data Is No Longer Enough
Resilience Is More Than Strength
Designing for Uncertainty and Risk
Climate, Water, Heat and Changing Loads
Designing Infrastructure as a System
Failure, Redundancy and Cascading Effects
Adaptable and Repairable Engineering
How Engineers Should Make Future-Proof Decisions
The New Engineering Objective
References

Quick Links

Health and Fitness Tools

Engineering Calculators

Finance Calculators

Economics Calculators

Percentage Calculators

Sciences Calculators

Math Calculators

Unit Converters

Traditional engineering is built around a powerful idea: establish the conditions a structure is expected to encounter, calculate the resulting demands, provide adequate resistance, and verify that specified performance requirements are satisfied.

That approach remains fundamental. A bridge still needs adequate capacity. A retaining wall still needs stability. A drainage system still needs hydraulic capacity. A building still needs to resist gravity, wind, seismic and other applicable actions.

The problem is not that these principles are wrong.

The problem is that the future conditions used to define the design problem may themselves be wrong.

A road designed around historical traffic patterns may experience very different freight demand. A drainage system designed from historical rainfall statistics may encounter rainfall intensities outside the range suggested by the historical record. A coastal facility may face water levels that differ materially from those assumed when it was designed. A building may remain structurally adequate while its mechanical systems, electrical equipment or access routes become unusable.

This creates a fundamental engineering distinction:

A structure can be correctly designed for its assumed future and still perform poorly in the actual future.

That is the problem resilience engineering attempts to address.

Recent NIST work on forward-looking building and infrastructure standards explicitly identifies future hazards, nonstationary reliability, changing load effects, material deterioration, adaptive design and resilience as areas requiring improved engineering methods.

The Problem With Designing for a Fixed Future

Every engineering design contains assumptions about the future. A design life might be 50, 75 or 100 years. Yet engineers rarely know with precision what the environment, demand, surrounding development, technology or operational requirements will look like several decades later. The conventional design process therefore uses a design basis.

It may contain assumptions about:

design loads;
return periods;
traffic volumes;
rainfall;
temperature;
groundwater;
material properties;
soil parameters;
occupancy;
building use;
deterioration;
maintenance;
operational procedures.

These assumptions are necessary. Engineering cannot proceed without them. But they should not be confused with certainty.

The traditional design chain

Traditional design question	Hidden assumption
What is the design load?	Future load can be adequately characterized
What is the design flood?	Historical or modeled statistics represent future conditions
What traffic volume should be used?	Demand will remain within forecast bounds
What material strength is required?	Material properties and deterioration are sufficiently predictable
What design life is required?	Future environmental exposure is reasonably stable
What failure probability is acceptable?	The consequences of failure are adequately understood
What capacity is sufficient?	The infrastructure will continue serving approximately the intended function

The difficulty becomes particularly important when the underlying processes are nonstationary. A stationary statistical assumption means, broadly, that the statistical characteristics of a process remain sufficiently stable over time for the analysis being performed. But infrastructure is increasingly exposed to changing conditions. The issue is not simply that extreme events occur. Extreme events have always occurred.

The engineering problem is that the probability distribution used to characterize future conditions may itself change. NIST’s 2026 forward-looking design research specifically identifies nonstationary reliability and changes in hazard demand, structural capacity and geotechnical conditions as research priorities. This changes the design question from: What is the correct design value?

toward: How sensitive is the infrastructure to being wrong about the design value?

That is a much more consequential question.

Why Historical Data Is No Longer Enough

Historical observations remain indispensable. Engineers should not abandon measured data, established standards or probabilistic analysis. But historical data describes what happened, not necessarily what will happen throughout the entire service life of a new asset.

Consider a drainage structure. A conventional hydraulic design might estimate a design discharge from rainfall and catchment characteristics, then size a culvert or stormwater system accordingly. But several variables can change simultaneously:

$Q=f(P,C,A,L,\ldots)$

where:

$Q$ = runoff or design discharge;

$P$ = precipitation characteristics;

$C$ = runoff coefficient;

$A$ = catchment characteristics;

$L$ = land-use conditions.

Urbanization can increase impervious area. Deforestation can alter runoff response. Development upstream can change drainage pathways. Climate conditions can alter precipitation intensity. A drainage system can therefore experience a future different from the one represented by the historical hydrologic record.

FHWA research has explicitly examined the sensitivity of transportation drainage infrastructure—including culverts, roadside drainage, storm drains and bridges—to extreme precipitation, streamflow and changing conditions. The same principle applies elsewhere.

Infrastructure variables that can change

Infrastructure element	Future variable	Possible engineering consequence
Road	Traffic and axle loading	Faster pavement deterioration
Bridge	Flood level/scour	Foundation vulnerability
Drainage	Rainfall intensity	Overtopping or flooding
Building	Occupancy/use	Changed loads and egress requirements
Coastal facility	Water levels	Flood exposure
Pavement	Temperature	Rutting, softening or accelerated deterioration
Concrete structure	Exposure environment	Durability and reinforcement deterioration
Power system	Heat and demand	Reduced equipment performance and increased demand
Water network	Population growth	Capacity deficiency
Tunnel	Extreme rainfall	Flooding and operational interruption

This is why climate-resilient infrastructure cannot simply mean “design for a larger number.” The problem is multidimensional. OECD research notes that transport, energy, telecommunications and water infrastructure need adaptation to changing climate conditions and evolving extreme-event risks.

Resilience Is More Than Strength

One of the most important conceptual changes in modern infrastructure engineering is the distinction between resistance and resilience. A strong structure is not automatically a resilient system. Suppose two bridges experience the same extreme event.

Bridge A suffers modest structural damage but remains closed for six months because critical components cannot be procured.

Bridge B suffers somewhat greater localized damage but can be inspected, repaired and reopened within three weeks.

Which bridge is more resilient? The answer is not determined by peak structural capacity alone. A useful conceptual measure is the functionality of the system over time:

$R=\frac{\int_{0}^{T}Q(t)\,dt}{Q_0T}$

where:

$R$ = normalized resilience measure;

$Q (t)$ = infrastructure functionality at time $t;$

$Q_{0}$ = pre-event functionality;

$T$ = evaluation period.

This is not a universal code equation; it is a useful engineering representation of the underlying idea. The important feature is that resilience considers the entire performance trajectory, not only the maximum load resisted.

Conceptual resilience curve

Stage	Infrastructure condition
Normal operation	100% functionality
Hazard occurs	Functionality decreases
Immediate response	Damage is assessed and isolated
Stabilization	Critical services maintained
Repair	Functionality progressively restored
Recovery	Normal or acceptable service resumes

NIST’s resilience research similarly describes infrastructure resilience in terms that include resistance to hazard effects, limiting failure, adaptation to future conditions and time to recovery of function.

This leads to a broader engineering objective:

$\boxed{\text{Resilience}= \text{Resistance}+ \text{Adaptability}+ \text{Recoverability}}$

A resilient asset therefore does not necessarily remain undamaged.It is designed so that damage does not automatically become loss of essential function.

Designing for Uncertainty and Risk

Uncertainty is not an engineering failure. Ignoring uncertainty is. Every design contains uncertainties in loads, resistance, deterioration, construction quality, models and future conditions.

A simplified reliability representation is:

$P_f=P[g(X)\leq0]$

where

Pf is the probability of failure and $g(X)$

g(X) represents the limit-state function.

But resilience introduces another dimension. Two systems can have similar probabilities of failure while having radically different consequences.

Therefore:

$\boxed{\text{Risk}=\text{Probability}\times\text{Consequence}}$

In practice, consequence can include:

loss of life;
economic loss;
service interruption;
environmental damage;
emergency response difficulty;
network disruption;
repair cost;
social disruption;
secondary infrastructure failures.

NIST research on risk-informed infrastructure decisions emphasizes that conventional approaches can be vulnerable to rare unforeseen events and recommends life-cycle risk-informed decision-making that considers sustainability, resilience and adaptive capacity.

This changes how alternatives should be compared.

Example

Suppose an engineer has three flood protection options:

Option	Initial cost	Expected damage	Recovery time	Adaptability
A	Low	High	Long	Low
B	Medium	Medium	Moderate	Moderate
C	High	Low	Short	High

Selecting A solely because it has the lowest initial construction cost is not necessarily economically rational.

The proper comparison should consider:

$C_{\text{life cycle}} = C_{\text{initial}} + C_{\text{maintenance}} + C_{\text{failure}} + C_{\text{downtime}} + C_{\text{adaptation}}$

The exact formulation depends on the project, but the principle is fundamental:

The cheapest structure to build is not necessarily the cheapest infrastructure to own.

Climate, Water, Heat and Changing Loads

Climate change is one of the clearest reasons why infrastructure design must increasingly consider changing future conditions. But resilience engineering should avoid a simplistic interpretation such as: Climate change means increase every design load.

That is not technically sufficient. Different hazards respond differently, and uncertainty differs between locations, variables and time horizons. NIST’s current forward-looking standards research specifically identifies the need to address uncertainty in hazard projections, local downscaling and conversion of projected hazards into engineering design criteria.

A more defensible design process

Instead of selecting one supposedly perfect future:

$\text{Future} = \text{one forecast}$

engineers can examine:

$\text{Future conditions} = \{S_1,S_2,S_3,\ldots,S_n\}$

where each Si represents a plausible future scenario.

The design can then be tested against each scenario.

Example: drainage infrastructure

Suppose:

Scenario A: moderate rainfall increase;
Scenario B: substantial rainfall increase;
Scenario C: rainfall increase + urbanization;
Scenario D: rainfall increase + urbanization + downstream flood constraint.

The resilient solution is not necessarily the one that maximizes pipe diameter.

It may instead combine:

additional conveyance;
detention;
overflow routes;
protected critical equipment;
flood-compatible access;
monitoring;
emergency operating procedures;
space for future expansion.

That is adaptation by design.

USACE already incorporates projected sea-level change into planning, engineering, design, construction, operation and maintenance through risk-based guidance and scenario tools.

Transportation engineering provides another example. FHWA’s resilience programs address flooding, wildfire, extreme weather, sea-level rise and other changing conditions rather than treating resilience as purely a structural-strength issue.

Designing Infrastructure as a System

A major weakness in conventional engineering is the tendency to evaluate assets independently. A bridge engineer designs a bridge. A drainage engineer designs drainage. A power engineer designs the electrical system. A communications engineer designs communications.

But infrastructure users experience the system, not individual engineering disciplines. A bridge may remain structurally safe while the road approaching it is flooded. A hospital may remain structurally intact while electrical power is unavailable. A pumping station may survive flooding while the electrical substation supplying it fails. A road network may remain physically intact while a critical bridge closure causes severe regional congestion.

This is the problem of interdependency. FEMA identifies infrastructure dependencies as an important component of community risk and resilience, while NIST research emphasizes that buildings and infrastructure systems need to consider interdependency and recovery of function.

Simplified infrastructure dependency chain

Extreme Event
      ↓
Electrical Substation Failure
      ↓
Power Loss
      ↓
Water Pump Failure
      ↓
Water Supply Reduction
      ↓
Hospital / Industry / Residential Disruption
      ↓
Economic and Social Consequences

The initiating event may therefore be relatively localized. The consequences may not be. This means resilience assessment should ask: What does this asset depend on, and what depends on this asset? That question can reveal vulnerabilities that ordinary component-level structural calculations cannot.

Failure, Redundancy and Cascading Effects

Engineering traditionally tries to prevent failure.

Resilience engineering also asks: If something fails, what happens next?

This is a fundamentally different question.

Consider a network containing components A, B, C and D. If failure of A causes B to overload, and B’s failure disconnects C and D, the original failure has propagated. The system has low tolerance for cascading failure.

Redundancy changes the failure path

Single pathway:

A → B → C → Service


Redundant pathway:

       → B →
A ─────────────→ Service
       → C →

Redundancy can involve:

alternative routes;
parallel power supplies;
multiple drainage paths;
emergency generators;
duplicate communications;
multiple load paths;
compartmentation;
alternative water sources;
replaceable structural components.

But redundancy has a cost.

The engineering problem is therefore not: Add redundancy everywhere.

It is: Identify where loss of redundancy creates unacceptable consequences.

This is where criticality analysis becomes important.

ASCE’s 2026 policy statement on resilient infrastructure specifically supports multihazard risk assessment, performance measures, infrastructure interdependencies, minimum performance goals and return-to-service planning.

The principle is straightforward:

$\text{Criticality} \approx \text{Consequence of Loss of Function}$

A small drainage culvert and a major emergency-access bridge should not necessarily receive identical resilience objectives.

Adaptable and Repairable Engineering

There is another assumption hidden inside conventional design: The structure we build today will be used in approximately the same way throughout its design life.

That assumption is increasingly weak. Buildings change occupancy. Roads experience changing traffic. Industrial facilities change processes. Transit systems change operating patterns. Technology changes equipment requirements. Urban density changes demand.

Consequently, a resilient asset should sometimes be designed not merely to survive future conditions but to change with them.

Examples of adaptability

Design strategy	Engineering purpose
Oversized structural provisions	Allow future changes in loading
Reserved service corridors	Permit future utilities
Modular components	Simplify replacement
Accessible critical equipment	Reduce repair time
Replaceable sacrificial elements	Localize damage
Expansion space	Accommodate future capacity
Multiple drainage pathways	Provide failure tolerance
Monitoring systems	Detect deterioration before critical failure
Flexible building layouts	Accommodate changing use
Protected foundations	Permit future environmental adaptation

Repairability deserves particular attention. A structure can be extremely strong but difficult to repair.

Conversely, a system with controlled, accessible and replaceable failure mechanisms may recover much faster. NIST’s work on functional recovery explicitly focuses on designing buildings and lifeline systems around recovery objectives rather than only preventing damage.

This introduces a useful design sequence:

$\boxed{ \text{Prevent} \rightarrow \text{Limit} \rightarrow \text{Isolate} \rightarrow \text{Repair} \rightarrow \text{Recover} }$

The objective is not to guarantee that nothing breaks. The objective is to ensure that what breaks does not determine the fate of the entire system.

How Engineers Should Make Future-Proof Decisions

The practical answer is not to abandon conventional design codes. Codes remain essential because they establish minimum requirements, common terminology, accepted methods and safety margins.

The improvement is to treat code compliance as the foundation of resilience—not the complete definition of resilience.

NIST’s assessment of existing codes and standards found that resilience involves hazard characterization, expected performance, recovery of function, interdependencies and changing environmental conditions.

A more robust engineering workflow can therefore be structured as follows.

Resilient design workflow

Step	Engineering question
1. Define function	What must the infrastructure continue to do?
2. Identify hazards	What can disrupt that function?
3. Identify uncertainty	Which assumptions are least certain?
4. Identify dependencies	What external systems are required?
5. Establish performance targets	How much degradation is acceptable?
6. Test scenarios	What happens under plausible future conditions?
7. Identify failure modes	How can local damage propagate?
8. Evaluate consequences	What happens if the system stops functioning?
9. Compare interventions	Which measures reduce risk most effectively?
10. Plan recovery	How quickly can essential service return?
11. Monitor	What evidence will indicate changing conditions?
12. Adapt	What can be modified when assumptions become obsolete?

This is fundamentally different from trying to produce one perfect forecast. It is closer to engineering against uncertainty. A particularly powerful concept is the adaptation trigger. For example:

If measured flood levels exceed a specified threshold for a defined frequency, initiate drainage expansion.

Or:

If traffic demand reaches a predetermined percentage of design capacity, implement the next capacity intervention.

Or:

If structural deterioration exceeds a defined condition state, initiate strengthening.

This creates a feedback loop:

$\text{Design} \rightarrow \text{Monitor} \rightarrow \text{Compare} \rightarrow \text{Adapt}$

The infrastructure therefore becomes part of an ongoing engineering management system, rather than a static object designed once and forgotten.

The New Engineering Objective

The central question was:

Should engineers optimize structures for the most likely future—or for the consequences of being wrong?

The technically defensible answer is:

Engineers should use the most credible information about the likely future, but optimize critical infrastructure against the consequences of uncertainty rather than against a single forecast.

That distinction matters. Designing only for the most likely future can produce an economically efficient solution if the forecast is approximately correct.

But if the consequences of being wrong are catastrophic, the expected value of a slightly more robust solution can be much greater.

This can be represented conceptually as:

Preferred Design = arg min [C_life cycle + C_failure + C_downtime + C_adaptation]

subject to acceptable:

$P_f$

and required:

Performance

The exact optimization will differ by project, but the engineering logic is universal.

The Fundamental Shift:

Traditional emphasis	Resilient engineering emphasis
Design load	Range of plausible demands
Design life	Life-cycle performance
Failure prevention	Failure consequence management
Component capacity	System functionality
Single hazard	Multihazard exposure
Static design	Adaptive design
Strength	Strength + ductility + redundancy + recoverability
Initial cost	Life-cycle cost
Historical conditions	Historical + projected conditions
Probability of failure	Probability × consequence
Construction completion	Long-term performance
Design once	Monitor, reassess and adapt

The engineering profession is therefore not moving away from prediction. Prediction remains essential. What is changing is the recognition that prediction has limits.

A resilient bridge is not one that assumes flooding will never exceed its design event. It is one in which critical failure modes have been understood, vulnerable components are protected or replaceable, alternative routes exist where justified, inspection and repair can be performed efficiently, and the consequences of disruption have been considered.

A resilient drainage network is not necessarily one with infinitely large pipes. It may be a system in which exceedance flows have safe pathways rather than being allowed to become uncontrolled urban flooding.

A resilient building is not necessarily one that remains completely undamaged. It may be one that protects critical structural and nonstructural components, limits disproportionate damage, preserves essential functions and permits rapid repair.

And a resilient infrastructure network is not one that never fails. It is one that fails in controlled ways, limits cascading consequences, preserves critical functions and recovers before disruption becomes systemic.

That is the deeper change in engineering thinking.

The old question was:

What loads should this structure withstand?

The more complete question is:

What range of conditions could it encounter, how could it fail, what would happen if our assumptions were wrong, and how quickly could we restore the function that society actually depends upon?

The future cannot be predicted with certainty. But infrastructure can be designed so that uncertainty does not automatically become catastrophe. That may be the most important objective of resilient engineering.

Selected Engineering References

NIST — Forward-Looking Codes and Standards — 2026 synthesis covering future hazards, future load effects, design criteria, adaptation and resilience.
NIST — Forward-Looking Codes and Standards: Future Hazards — addresses uncertainty in hazard projections and conversion of projections into engineering design criteria.
NIST — Forward-Looking Codes and Standards: Engineering Design Guidelines and Criteria — covers nonstationary reliability, service life, design loads and degradation of structural capacity.
NIST — Assessment of Resilience in Codes, Standards, Regulations, and Best Practices — examines performance, recovery, interdependencies and changing environmental conditions.
NIST — Risk-Based Decision Making for Sustainable and Resilient Infrastructure — life-cycle and risk-informed infrastructure decision-making.
FHWA — Infrastructure Resilience to Extreme Events and Climate Change — transportation infrastructure sensitivity to precipitation, streamflow and climate-related extremes.
FHWA — Extreme Events and Resilience Resources — federal guidance and technical resources for extreme-event resilience.
USACE — Sea-Level Change — risk-based incorporation of projected sea-level change into infrastructure planning and engineering.
OECD — Infrastructure for a Climate-Resilient Future — infrastructure adaptation and resilience under changing climate risks.
ASCE — Resilient Infrastructure Policy Statement — emphasizes multihazard risk assessment, interdependencies, performance objectives and return-to-service planning.